20 open-source tools for AI agent security, organized into 6 categories
MaryamMiradi · x · 2026-10-11
The author outlines the security concerns when building AI agents that browse the web, call tools, execute code or touch internal systems — prompt injection, excessive permissions, unsafe tool use, leaked secrets — and curates 20 open-source tools by scenario:
- Injection & guardrails: LLM Guard, NeMo Guardrails, AgentGuard, Guardian
- Red teaming: PyRIT, garak, Promptfoo, Giskard
- Identity & permissions: Open Policy Agent, OpenFGA, Casbin, Cedar
- Sandboxing & runtime: E2B, gVisor, Firecracker, Kata Containers
- Data & secrets: Presidio, Trivy, Semgrep, ModelScan
Key takeaway: production agent security goes far beyond prompt injection. A benchmark comparing several of these tools may follow.
More from coding & agent
- Nous Research launches stealth coding and agentic reasoning model, free for limited time — Teknium · 2026-10-11
- Cloud agent users mock devs still lugging around laptops in viral meme — steipete · 2026-10-11
- MCP server plus a World of Warcraft environment: AI agents entering Azeroth? — djcows · 2026-10-11
- Using two cleanup agents on a Mac to tame multi-agent RAM and storage bloat — pvncher · 2026-10-11
- Dev lets an agent read and answer his email, soliciting tests until deadline — DanielLockyer · 2026-10-11
- Haiku fails 15 of 17 agent runs on output format, deemed no upgrade over Luna — zeeg · 2026-10-11